使用Intel OpenVINO Model Optimizer转换TensorFlow模型报错求助

问题详情 #

我用以下代码保存图像分类模型：

import tensorflow as tf
model = tf.keras.models.load_model('model.h5')
tf.saved_model.save(model,'model')

环境信息 #

• Google Colab上TensorFlow版本：2.9.2
• 本地Intel OpenVINO开发工具版本：2021.4.2 LTS

执行命令及报错 #

运行Model Optimizer命令后出现框架错误，命令及完整输出如下：

C:\Program Files (x86)\Intel\openvino_2021.4.752\deployment_tools\model_optimizer>python mo_tf.py --saved_model_dir C:\Users\dchoi\CNNProejct_Only_saved_English\saved_model --input_shape [1,32,320,240,3] --output_dir C:\Users\dchoi\CNNproject_only_output_English\output_model
Model Optimizer arguments:
Common parameters:
        - Path to the Input Model:      None
        - Path for generated IR:        C:\Users\dchoi\CNNproject_only_output_English\output_model
        - IR output name:       saved_model
        - Log level:    ERROR
        - Batch:        Not specified, inherited from the model
        - Input layers:         Not specified, inherited from the model
        - Output layers:        Not specified, inherited from the model
        - Input shapes:         [1,32,320,240,3]
        - Mean values:  Not specified
        - Scale values:         Not specified
        - Scale factor:         Not specified
        - Precision of IR:      FP32
        - Enable fusing:        True
        - Enable grouped convolutions fusing:   True
        - Move mean values to preprocess section:       None
        - Reverse input channels:       False
TensorFlow specific parameters:
        - Input model in text protobuf format:  False
        - Path to model dump for TensorBoard:   None
        - List of shared libraries with TensorFlow custom layers implementation:        None
        - Update the configuration file with input/output node names:   None
        - Use configuration file used to generate the model with Object Detection API:  None
        - Use the config file:  None
        - Inference Engine found in:    C:\Users\dchoi\AppData\Local\Programs\Python\Python38\lib\site-packages\openvino
Inference Engine version:       2021.4.0-3839-cd81789d294-releases/2021/4
Model Optimizer version:        2021.4.2-3974-e2a469a3450-releases/2021/4
[ WARNING ] Model Optimizer and Inference Engine versions do no match.
[ WARNING ] Consider building the Inference Engine Python API from sources or reinstall OpenVINO (TM) toolkit using "pip install openvino==2021.4"
2022-11-19 01:34:44.207311: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2022-11-19 01:34:44.207542: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
C:\Users\dchoi\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\autograph\impl\api.py:22: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
  import imp
2022-11-19 01:34:46.961002: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set
2022-11-19 01:34:46.961949: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found
2022-11-19 01:34:46.962904: W tensorflow/stream_executor/cuda/cuda_driver.cc:326] failed call to cuInit: UNKNOWN ERROR (303)
2022-11-19 01:34:46.969471: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host: DESKTOP-SCBPOUA
2022-11-19 01:34:46.969727: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname: DESKTOP-SCBPOUA
2022-11-19 01:34:46.970663: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-11-19 01:34:46.971135: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set
[ FRAMEWORK ERROR ]  Cannot load input model: SavedModel format load failure: NodeDef mentions attr 'validate_shape' not in Op<name=AssignVariableOp; signature=resource:resource, value:dtype -> ; attr=dtype:type; is_stateful=true>; NodeDef: {{node AssignNewValue}}. (Check whether your GraphDef-interpreting binary is up to date with your GraphDef-generating binary.).

已尝试运行install_prerequirement/install_prerequisites_tf2.bat，但报错依旧，寻求解决方法。

解决方法 #

1. 版本兼容性修复
   OpenVINO 2021.4仅支持TensorFlow 2.5.x及以下版本，你使用的TensorFlow 2.9.2超出兼容范围，导致AssignVariableOp节点属性不匹配。

2. 降级TensorFlow重新导出模型
   在Colab中切换到TensorFlow 2.5版本，重新加载并导出SavedModel：

   !pip install tensorflow==2.5.0
   import tensorflow as tf
   model = tf.keras.models.load_model('model.h5')
   tf.saved_model.save(model, 'tf25_model')
   

   将新导出的tf25_model文件夹下载到本地，重新运行Model Optimizer命令。

3. 升级OpenVINO版本
   若不想降级TensorFlow，可升级OpenVINO到2022.1及以上版本（该版本开始支持TensorFlow 2.8+），确保工具链版本与TensorFlow匹配。

4. 修复版本不匹配警告
   先按提示重新安装对应版本的OpenVINO Python API：

   pip install openvino==2021.4
   

内容的提问来源于stack exchange，提问作者newbieLife